cpp is a ggml-based transcription library designed to support a wide range of modern automatic speech recognition (ASR) models. The tool addresses the challenge of distributing cross-platform speech-to-text applications by providing a fast and accurate inference engine that works on Mac, Windows, and Linux. It was developed to offer a reliable solution for local speech-to-text tasks, particularly for developers seeking to embed ASR capabilities in desktop or mobile applications without relying on large dependencies.
The library supports 16 ASR model families, encompassing over 60 models, and aims to expand its compatibility with additional state-of-the-art transcription models. Each supported model has undergone numerical validation and word error rate (WER) testing to ensure output matches the reference implementation. Performance is a core focus, with acceleration available through Vulkan, Metal, CUDA, and TinyBLAS. The tool offers both streaming and batch transcription modes, catering to diverse use cases.
cpp is written in C/C++ and provides official language bindings for Python, JavaScript/TypeScript, Rust, and Objective-C/Swift, making it accessible for integration in a variety of development environments. cpp, and is intended to be easily embeddable without the overhead of larger machine learning frameworks like PyTorch. The library also publishes benchmark results for supported models and shares numerical validation data in its repository and on Hugging Face.
The tool is maintained by the author of Handy and was created out of the need for a trustworthy, high-performance ASR library with broad model support and robust distribution. Its development is driven by practical experience supporting a cross-platform ASR application and by feedback from a wide range of ASR use cases.
In the Voice, TTS & speech space, Project takes a focused approach. Enabling developers to perform accurate, fast speech-to-text transcription using modern models locally. Project is an open-source project aimed at developers building speech-to-text applications. The project is open source (MIT). The product ships for the web and the command line, and it can be self-hosted.
Behind Project is Handy (author/maintainer), and the product first shipped in 2026. The project is developed in the open on GitHub with 21 stars and 473 commits in the last 90 days. Among its 5 catalogued features are speech-to-text, model validation, and cross-platform.
Latest indexed changes and source events
Transcribe.cpp – ggml based transcription engine verified by the PulseGate indexer
Other apps tracked under the same category.
